Conflict Environments and Civil War
Rachel Myrick
Lindsay Reid
Stanford University
University of North Carolina, Chapel Hill
[email protected]
[email protected]
Kelly M. Kadera
Mark J.C. Crescenzi
University of Iowa
University of North Carolina, Chapel Hill
[email protected]
[email protected]
1
June 25, 2015
1 Replication
files for this research can be found at http://dx.doi.org/10.7910/DVN/NDZV42
Abstract
In this paper we demonstrate that a state’s conflict environment a↵ects its proclivity for
experiencing civil war. We conceptualize a state’s conflict environment as including spatial
and temporal dimensions of conflict di↵usion. As conflict in one’s environment becomes
more spatially and temporally proximate, the likelihood of civil war onset increases. Based
on theoretical expectations, we construct the Conflict Environment (CE) score, a composite
indicator that taps into the spatially and temporally proximate violence in a state’s neighborhood. The constitutive elements of the CE score decay across both geographic distance
and time. We incorporate the CE score into the standard models of civil war onset and
demonstrate that the conflict environment is a robust determinant of civil war, even when
domestic factors are taken into account. The newly developed measure is flexible and customizable, and in its current form, allows us to pinpoint the international dynamics of civil
war.
Word Count: 9,425
Introduction
Can a state’s surrounding environment catalyze the onset of civil war? Analyses of civil war
onset often focus on domestic explanations, demonstrating that long standing grievances
over economic disparities, discrimination, access to basic government services, and political
rights spur rebellions. Beyond these domestic factors, scholars recognize that civil wars
do not occur in isolation of each other, but external influences are less well understood.
When armed combatants migrate from one state to another, when internal violence disrupts
regional economic stability, or when recent successes by rebels or secessionists embolden
others, civil war is also driven by regional and global factors.
Consider for example the 2012 insurgency in northern Mali, which has been unequivocally fueled by a history of internal conflict and unrest. Since independence, Tuaregs have
felt marginalized from the southern government in Bamako. Vying for their own independence, they organized an unsuccessful uprising against the Malian government in 1962. Over
the subsequent decades, Malian Tuaregs incited rebellions in 1990 and 2007 in response to
political marginalization and socioeconomic hardships. In recent years, drought and food
shortages have exacerbated social tensions.
However, neither the magnitude nor the timing of Mali’s civil conflict can be divorced from
its broader regional context. The 2012 conflict has direct links to the Libyan Civil War that
culminated in late 2011. In a Security Council briefing at the United Nations in January 2012,
Ambassador Rosemary DiCarlo anticipated these links, remarking, “We recognize that the
Libyan crisis has brought a new set of cross-border challenges relating to security, including
increased illicit weapons trafficking that pose a threat to the stability of the region.” DiCarlo
further expressed concern that such weapons “could further destabilize already fragile areas
of the Sahel and surrounding regions” (DiCarlo January 26, 2012). The porous borders in
the Sahel region enabled migrants and arms to travel long distances. A report issued by the
UN Security Council after the overthrow of Muammar al-Gaddafi expressed concern with the
“influx of hundreds of thousands of traumatized and impoverished returnees as well as the
inflow of unspecified and unquantifiable numbers of arms and ammunition from the Libyan
arsenal” (Ki-Moon 2012, p. 2). The International Organization for Migration estimated
that 420,000 people fled Libya, 30,000 of which returned to Mali (Ki-Moon 2012, p. 6).
These migrants included Tuareg combatants who aided the Gaddafi regime and returned
to Mali trained and armed; the head of the MNLA, Mohammed Ag Najim, was a former
colonel in Gaddafi’s army.
Not only was Mali’s 2012 conflict spurred by spatial proximity to Libya, it was also
influenced at various points in the preceding decade by recent civil conflicts a✏icting Algeria,
Niger, Sierra Leone, Liberia, Côte d’Ivoire, Senegal, Guinea, and Nigeria. The temporal
proximity of conflict in Mali’s neighborhood reinforced feelings of insecurity and dominated
the consciousness of actors within the country. Mali’s conflict environment – a spatial and
temporal context of insecurity and violence – serves as a reminder that external factors
contribute to the onset of conflict.
Research on the e↵ects of nearby violence, however, has focused largely on the dimension
of space. Studies of conflict di↵usion often rely on contiguity or short temporal lags. The
result is a gap in our ability to understand how conflict environments trigger civil wars.
In this paper, we develop a model of a state’s conflict environment to facilitate a better
understanding of how a state’s neighborhood exacerbates the risk of civil war. Building
on existing di↵usion theories, we develop a conflict environment measure that represents
both spatial and temporal dimensions of neighborhood violence. In so doing, we address
how and why the conflict environment matters for states while at the same time recognizing
the causal complexity of conflict. Our conflict environment approach to understanding the
causes of civil war heeds Solingen’s (2012) advice to “integrate domestic, regional, and global
considerations under a common theoretical framework” (p. 640). Modeling exogenous factors
is a complement to, not a substitute for, understanding domestic determinants of conflict.
2
What We Already Know About Exogenous Influences on Internal
Conflict
Previous analyses have examined regional and global trends to consider how these external
factors influence the occurrence of conflict within a state’s borders. On a systemic level,
the relationship between international structure and the occurrence of civil war has received
considerable attention following the demise of the U.S.S.R. Lacina (2004), for example,
argues: “The end of the Cold War fundamentally altered the place of civil war in international
politics” (191). She suggests that the changes in the polarity of the international system and
the resurgence of nationalism following the dissolution of the Soviet Union led to the rapid
rise in internal armed conflict during the 1990s.
An additional body of literature focuses on exogenous determinants of civil conflict below the systemic level. Gleditsch (2002), for instance, encourages researchers to “study
conflict, cooperation, and democratization through regional interactions rather from more
conventional perspectives that emphasize relations at the level of either the global system or
individual states” (p. 10). This work recognizes that civil wars, while domestically driven,
do not occur within a vacuum. In order to gain a full picture of the causes and risk factors
for civil war, one must look at what is going on in a state’s neighborhood. This literature
lays a foundation for our own theoretical and empirical approach, which couples exogenous
regional forces with recognized domestic determinants of conflict.
A closer look at contagion and clustering
Early conflict di↵usion research focused on interstate war and conceptualized it as a disease
that infects neighboring states (Alcock 1972; Houweling and Siccama 1985). Most and
Starr (1980) suggest that states with neighbors involved in interstate conflict are more likely
to engage in warfare than states with no neighbors engaged in conflict. Siverson and Starr
(1990) highlight how both geographic (borders) and political (alliances) factors increase the
probability that a nation would go to war. Taking a more theoretical approach, Kadera
3
(1998) focuses on the mechanisms and barriers to the transmission of conflict. Transmission
mechanisms (such as defense pacts) enable the spread of violence, while barriers (such as
distance) and resource constraints diminish contagion.
Contemporary studies of civil war di↵usion draw on arguments similar to those in the
earlier interstate war di↵usion literature, however, empirical evaluations of civil war onset
provide mixed support, at best, for the contagion phenomenon. In an analysis of fortyseven civil war onset studies, Dixon (2009) demonstrates the contradictory findings about
the e↵ect of a state’s neighborhood on its own propensity for civil conflict. Using regional
dummy variables to assess the impact of neighborhoods on civil conflict, Fearon and Laitin
(2003) find that only one geographic region (Asia) has a significant e↵ect on civil war onset
and ultimately exclude regional variables from their baseline model. Hegre and co-authors
(2001) also find neighboring civil war to be an insufficient explanation for conflict outbreak.
On the other hand, Sambanis (2001) finds a positive correlation between neighboring ethnic
war and war onset. Similarly, Hegre and Sambanis (2006) determine that the existence of
war-prone neighbors is a determinant of civil war onset.
In perhaps the most comprehensive study of civil war di↵usion, Buhaug and Gleditsch
(2008) confirm that conflict in a contiguous state increases the likelihood of civil war onset in the focal state. However, more nuanced measures of neighboring conflict perform
poorly. Buhaug and Gleditsch conclude that “neither the distance to the nearest conflict,
the weighted density of conflict in the neighborhood, the influx of refugees from a conflict
neighbor, nor the severity of the neighboring conflict explains the trajectory of contagion”
(230). Their discussion highlights one particular di↵usion mechanism—transnational ethnic
ties between states. Other researchers have since investigated this link, providing compelling
evidence that transborder kin in conflict make civil war more likely in a focal state (Cederman, Girardin and Gleditsch 2009; Cederman et al. 2013; Forsberg 2014b).
One probable reason for a lack of scholarly consensus on the existence and prevalence of
conflict di↵usion is that researchers are divided as to whether the spatial clustering of civil
4
conflict is evidence of a contagion phenomenon (Salehyan and Gleditsch 2006; Buhaug and
Gleditsch 2008; Braithwaite 2010a; Forsberg 2014b) or simply an artifact of clustered states
with similar domestic attributes (Hegre et al. 2001; Elbadawi and Sambanis 2002; Gleditsch
2002; Fearon and Laitin 2003; Collier and Hoe✏er 2004; Bosker and de Ree 2014). For
instance, Gates and his colleagues (2006) investigate the clustering of political characteristics,
demonstrating that relatively homogenous political neighborhoods experience more stability.
In short, the character of external influences has been under-explored. Scholars must think
critically about the transmission mechanisms involved in the di↵usion process and create
models that appropriately capture these dynamics.
The Mechanisms of Conflict Di↵usion
How do states’ environments influence their own likelihood of experiencing civil war? Here,
we propose an explanation that addresses the complex contagion dynamics by framing the
environmental e↵ect in terms of the spatial and temporal proximity of others’ civil wars.
This approach allows us to incorporate both direct and indirect di↵usion mechanisms and to
model some of the regional risk factors that produce similar e↵ects for neighboring states.
Direct and indirect mechanisms both drive the spread of conflict. Direct (or physical)
di↵usion happens when geographically proximate violence, especially when it is in contiguous
states, produces externalities. For example, Murdoch and Sandler (2002) discuss spillover
economic e↵ects of violence including damage to infrastructure, disturbance of trade patterns,
and withdrawal of foreign direct investment from a region. These e↵ects, in turn, spur
political unrest in a fragile neighboring state. The movement of weapons and ammunition
from one state’s conflict also abets violence in nearby states. An influx of arms might
transform a simmering dispute into a full blown crisis. Finally, emigration of people (whether
refugees or trained combatants) across porous borders may generate new waves of violence
by fostering economic instability (Salehyan 2008) or by altering the domestic power balance
(Krcmaric 2014). In the latter scenario, a demographic shift undermines an existing balance
5
between political groups, making it difficult for them to credibly commit to nonviolence.
Negative externalities a↵ect not only nearby states, but entire neighborhoods. Severe
conflict triggers broad, regional aftershocks (Bremer 1982), producing spatially clustered
states with similar domestic risk factors such as weakened economies. Such e↵ects are not
mere artifacts of coincidence; they emerge from shared exposure to hostile environments.
Conflict may also indirectly a↵ect entire neighborhoods as well. Information about a
neighboring conflict, for example, can cause a rebel group to revise its beliefs about success (See, e.g. Hill, Rothchild and Cameron 1998). One such form of learning is through
demonstration e↵ects, whereby tactical successes incentivize neighboring groups to adopt
similar strategies (Kuran 1998). Rebels engaged in civil conflict recycle innovations used in
prior uprisings, fueling concentrated waves of protests and violence. Processes of learning
or imitation occur more readily when there is a preexisting link—such as ethnic, linguistic,
religious, or cultural ties—between a neighboring conflict and a focal state. For instance, the
presence of a separatist ethnic kin group in a neighboring country increases the likelihood of
secessionist movements (Saideman and Ayres 2000; Ayres and Saideman 2000).1
Rethinking time and space
Most cases of civil conflict di↵usion are probably best explained by a conjunction of direct and
indirect mechanisms. However, because the mechanisms associated with indirect di↵usion
processes are elusive, empirical studies of civil conflict onset often neglect them. In a recent
article, Forsberg (2014a) describes the difficulties of capturing indirect di↵usion processes in
1
Communication, especially via social media, is an important mechanism for di↵usion and
political contestation, but this process is not easily integrated into current spatial models.
Tufekci and Wilson (2012) provide an excellent introduction to the connection between social
media and protest. For those interested in developing measures of di↵usion driven by social
media or communication technology, we suggest developing them as complements to, rather
than replacements of, the current spatial approaches.
6
large-N studies:
“While spatial proximity may be relevant for direct forms of di↵usion, such a
restriction may be of less relevance when studying indirect di↵usion, as processes
of inspiration and strategic learning can travel longer distances. In addition,
the temporal dimension of di↵usion presents a challenge for future theoretical
development and associated statistical models. For instance, a standard time-lag
of any kind would miss several of the cases which regional experts would consider
to be di↵usion” (p. 195).
With this in mind, our analysis of civil conflict carefully considers both the spatial and
temporal components of a state’s neighborhood to better explain how regional context conditions domestic outcomes. In what follows, we present our conceptualization of the conflict
environment, a new means to incorporate diverse di↵usion mechanisms into models of civil
war onset.
While recent conflict literature demonstrates great progress in the analysis of geographic
space (Gleditsch and Ward 2001; Gleditsch 2007; Buhaug and Gleditsch 2008; Weidmann,
Kuse and Gleditsch 2010; Danneman and Ritter 2014), theoretical and empirical treatments
of the temporal impacts of civil conflict are less advanced. Scholars’ typical approach to
time has been to either set it aside in order to focus on other factors, or to use a simple
temporal lag (almost always in the form of one year). This approach hints at, but doesn’t
fully capture, the potential cumulative e↵ects of ongoing regional violence or peace. Here
we see an opportunity to improve the way we model the historical dynamics of the conflict
environment.
We propose a learning- and memory-based approach that considers a state’s unique
conflict history as generated by its own conflict experiences and those of its neighbors. We
eschew the standard time lags in favor of momentum e↵ects from recent and nearby violence
and decay e↵ects from long-ago and distant events. Thus, the e↵ects of environmental
instability do not switch o↵ after a set amount of time or because they are not immediately
7
next door. Instead, regional conflict lingers for years in a state’s collective memory and has
security consequences, albeit declining ones, over geographic space. Our approach uses three
simple assumptions.
First, when conflict occurs in a state’s local environment, the information it produces
immediately a↵ects actors within the state and then its valence degrades over time. Within
this framework, new conflict events take primacy in the information queue as old information
decays in importance. Thus, the impact of the conflict environment does not simply appear
and disappear, but rather fades over time.
Second, the speed at which conflict information decays is itself a function of relevant
histories of peace and violence. As years of neighborhood peace accumulate, memories
of violence fade more rapidly. If, on the other hand, the neighborhood is plagued with
numerous civil war events, this historical memory is harder to shake. A highly conflictual
neighborhood institutionalizes violence, making it harder to forgive and forget, increasing
the likelihood of conflict spillover, and priming actors to think of force as a legitimate form of
discourse. When a state’s or society’s consciousness is shaped by years of nearby instability,
the domestic environment in which decisions are made is much di↵erent, and much more
conflict-prone, than one that is dominated by the view that civil war is a rare and isolated
event. These claims are consistent with the logic of “conflict traps” and “conflict hot spots”
explored by Sambanis and his colleagues (2003) and Braithwaite (2010b), respectively.
Third, as a state experiences more civil war within its own borders, it becomes more sensitive to its conflict environment’s history. As a state’s domestic experiences become more
entrenched in internal violence, that violence becomes increasingly institutionalized, and the
decay of information contributing to its environmental history abates. When domestic tensions between factions endure, proximate and recent conflict in the region more persistently
condition the political, economic, and social systems within the focal state.
We have conceptualized a state’s conflict environment and laid out a variety of direct
and indirect mechanisms, operating over space and time, by which it augments a state’s
8
proclivity for civil war. Multiple mechanisms play a role in any particular case. In Mali, for
example, the cross-national spread of arms, combatants, and refugees all directly contributed
to conflict di↵usion, as did more indirect exposure to a recent history of civil war in other
African states. Our theoretical approach does not distinguish among these mechanisms.
Instead, it integrates them into a unified explanation of the kind of environmental conditions
that give rise to civil wars.
A more general link between a state’s environment and its susceptibility to civil war
can be found in the bargaining process that preserves peace among domestic actors. A
conflictual environment increases the likelihood of bargaining failures, thus increasing the
chances of civil war onset through two primary pathways. It amplifies commitment problems
given the uncertainty over changing conditions. Whether proximate conflict directly di↵uses
across borders or indirectly through mechanisms of learning and emulation, domestic actors
will have greater difficulty credibly committing to peaceful relations. The threat of changing
security conditions precludes the ability to commit to a peaceful bargain. Similarly, conflict in
the neighborhood may also lead to civil war onset insofar as it triggers authoritarian policies
and promulgates extremism, fueling tensions between opposition groups and governments.
Because a state’s conflict environment alters the ability to accurately evaluate intentions
and capabilities, the conflict environment drives bargaining failures, and ultimately, civil war
onset. Thus, we propose:
Hypothesis 1 As a state’s conflict environment worsens, it is more likely to experience the
onset of civil war.
Generating the Conflict Environment Score
To capture the background conditions against which the traditional domestic determinants
of civil war operate, we construct a Conflict Environment (CE) score. A state’s CE score
represents both spatial and temporal dimensions of neighborhood violence and help provide
a more holistic picture of how and why civil war occurs. We also note that the CE score
9
is customizable to other research agendas; raw conflict data can be drawn, for example,
from extant datasets on interstate wars, interstate militarized disputes, and civil wars. The
constituent lags can be built using either ongoing conflict, new conflict onsets, or conflict
intensity. As such, we hope that the CE score will serve as both a theoretical and empirical
contribution beyond its present application. For purposes of this paper, the core piece of the
CE score is civil conflict onset.
To construct the spatial component of the CE score, we begin with a standard N x N
matrix, where N is the total number of states in the system. Its cells are populated with
binary values, where 0 indicates that the row and column states are not neighbors, and 1
indicates that they are (Ward and Gleditsch 2008). Rather than only evaluating the impact
of contiguous states, our calculation includes larger geographic neighborhoods. Spatial diffusion, as the literature shows, does not require a shared border. Following convention, our
measure considers two states to be “neighbors” when the minimum distance between them
is 950 kilometers or less (Gleditsch and Ward 2001). We obtain data on minimum distances
between states from the CShapes dataset (Weidmann, Kuse and Gleditsch 2010). Following
Danneman and Ritter (2014), we replace the 0s and 1s in the matrix with distance-sensitive
weights; all states that share a “neighborhood” (i.e., dyads with a value of 1) receive a spatially lagged value reflecting their proximity. This spatial lag weights contiguous states more
heavily than states that are farther away. Each dyad’s cell in the matrix is generated with
this distance-degraded formula:
1
⇣ M inDistance ⌘ 14
950
(1)
We also set the diagonals of each weights matrix to 0, so conflict involving the focal state
will not be part of its own CE score. Finally, in order to capture the spatially-weighted
impact of nearby conflict, we multiply the row associated with the focal state by a vector
of annual conflict values across all states, resulting in a scalar value that is the distancedegraded spatial lag of conflict onset in the neighborhood a given country-year (slco). For
10
the purpose of this research, we use two di↵erent civil conflict vectors, one generated with
Intrastate War COW data and one generated with UCDP/PRIO Armed Conflict Data.2 We
also generate two spatial lags, one for ongoing conflict (slc) and one for conflict onset (slco).
After calculating the spatial components of the CE score, we add a temporal dimension.
The temporal lag for the model is based on the interstate interaction model developed by
Crescenzi and Enterline (2001). We adjust their functional form to make it consistent with
the theoretical assumptions outlined above. A state i ’s CE score in year t is given by:
CEit =
✓
e
1+ it
1+↵it
◆
CEit
1
+ slcoit
In this equation, CEit is a state’s CE score in a given year, CEit
(2)
1
is a state’s CE score
in the previous year, and slcoit is the spatial lag of civil conflict onset within the state’s
neighborhood at time t. The decay of neighborhood conflict’s e↵ect does not disappear after
one year, but dissipates over time. We capture the accumulation of continuous peace in
the neighborhood ( it ); more peace in the neighborhood accelerates the decay of old conflict
information. We represent the buildup of violent history within a state (↵it ) as a running
total of state i ’s civil conflicts. As such, a state’s history of armed conflict decelerates
this decay, making conflict memories more permanent. Table 1 illustrates the descriptive
statistics for civil CE scores across various datasets. As the values of the CE scores increase,
we expect civil war onset to become more likely.
The net result is a CE score that reflects both the spatial and temporal e↵ects of civil
conflict di↵usion from a state’s neighborhood to the state itself. Figure 1 plots annual CE
scores across all states in our analysis.
2
We have also generated CE measures using alternative civil war datasets from Fearon
and Laitin (2003) and Sambanis (2004) wars.
11
Figure 1: Conflict Environment Scores (1960-2006)
Table 1: Civil Conflict Environment Scores Across Civil War Data (1960-2006)
Statistic
Civil
Civil
Civil
Civil
CE
CE
CE
CE
Score
Score
Score
Score
(ACD v 4-2012)
(COW)
(Sambanis 2004)
(Fearon & Laitin 2003)
N
Mean
St. Dev.
Min
Max
5,971
5,602
5,208
5,024
0.142
0.248
0.126
0.087
0.334
0.529
0.337
0.266
0.000
0.000
0.000
0.000
3.291
9.646
3.713
3.000
Empirics: Embedding the CE Score in Standard Models
of Civil War Onset
One key advantage of our approach is that we craft the CE score such that it complements
current analyses of civil war onset. Incorporating environmental or extra-state factors often
leads scholars to tools such as network analysis, which can make comparisons to what we
already know about the onset of civil war difficult. Moreover, while Social Relations Models
12
(SRMs) and Exponential Random Graph Models (ERGMs) reduce bias in the estimation
of the central relationship of interest (e.g., GDP’s impact on civil war onset) and reveal
the data’s underlying structure, they do not theorize spatial dependencies and even newer
versions have difficulty managing time dependencies (Cranmer and Desmarais 2011; Dor↵
and Ward 2013). Our goal is to evaluate our hypothesis without dismissing widely recognized
causes of civil war. In this section, we therefore briefly review the state of the field in order to
develop the proper platform for our research design before laying out that design, conducting
the analyses, and performing robustness checks.
What we already know: Domestic explanations of civil war onset
Initial domestic explanations of conflict onset rest on economic motivations. Gurr’s theory of
relative deprivation, a grievance-based approach, describes how violent rebellion originates
from a perceived gap between a group’s expectations and its capabilities 1970. Inequalities across groups fuel mobilization (Stewart 2002). Studies have confirmed that unequal
resource distribution as well as political and socioeconomic disparities increase the risk of violent conflict (Regan and Norton 2005; Cederman, Gleditsch and Buhaug 2013). Economic
deprivation, low rates of economic growth, and low income levels are also empirically associated with civil war onset (Fearon and Laitin 2003; Hegre and Sambanis 2006). Poverty is
indicative of low state capacity, and therefore, an environment ripe for insurgency. Focusing
on economic opportunity for insurgency, Collier and his co-authors (Collier 2000; Collier
and Hoe✏er 2004; Collier, Hoe✏er and Rohner 2009) postulate that civil war is driven
by “greed.” Although grievances motivate rebels, groups must have sufficient opportunity to
pursue civil war in order for one to begin.
Beyond economic explanations of civil war, scholars have investigated the role of variables
such as size, state capacity, and regime type. Weak states, defined by their persistent inability
to deliver basic public services to a growing population, exacerbate collective grievances and
provide opportunity space for rebellion (Call 2011; Rotberg 2004). Regime type also has
13
implications for civil conflict. Highly institutionalized democracies and autocracies face a
very low risk of civil war, while semi-democracies and transitioning regimes are most likely
to experience conflict. Ellingsen and Gleditsch (1997) confirm the inverted U-shaped curve
that represents that relationship between regime type and risk of conflict onset. Hegre and
co-authors (2001) extend this analysis, arguing that regimes which fall somewhere in the
gray area between autocracy and democracy face a higher likelihood of civil war onset.
Finally, intuition suggests that ethnic and religious diversity determine civil war onset.
However, Fearon and Laitin (2003) argue, “It appears not to be true that a greater degree
of ethnic or religious diversity...by itself makes a country more prone to civil war” (75).
While this conclusion runs contrary to common views of ethnic divisions and conflict, further
analyses by Fearon, Kasara and Laitin (2007) and Hegre and Sambanis (2006) also find that
ethnic di↵erence is not robustly associated with civil war onset. Blimes (2006) attempts to
reconcile the disconnect between popular belief and scholarly works, and he finds an indirect
but statistically significant relationship between ethnicity and civil war onset. In what
follows, we include in our statistical models several variables to control for the determinants
of civil war that have been identified in existing research.
Research Design
To assess the extent to which the spatial and temporal treatment of a state’s conflict environment improves our understanding of internal armed conflict, we conduct a probit regression
analysis of civil war onset. Our universe of analysis includes all country-years from 1960
through 2006, so our observations are limited to the postwar and postcolonial era. We use a
binary dependent variable; country-years experiencing civil war onset are coded as 1 while
all others are coded 0.
As Hegre and Sambanis (2006) note, the coding for civil war onset is highly inconsistent
14
across various studies.3 To mitigate the e↵ects of these data discrepancies and ensure that
our findings are robust across definitions of civil war, we use two di↵erent civil conflict
datasets to code our dependent variable, one with intrastate war data from the Correlates of
War dataset (Sarkees and Wayman 2010; Singer and Small 1982) and one from the Uppsala
Conflict Data Project (Themner and Wallensteen 2012; Gleditsch et al. 2002), henceforth
COW and ACD, respectively.4 These datasets are widely used in studies of conflict behavior
and also represent the two most extreme thresholds for civil war coding. UCDP codes a
civil conflict using a minimum of 25 battle deaths while COW codes a civil war using a
minimum of 1000 battle deaths. The UCDP dataset further restricts armed conflicts to
those fought over government or territory, in which at least one party is the government of
a state (Gleditsch et al. 2002).
To control for domestic determinants of civil war, we first include a state’s per capita
income. As a proxy for economic well-being and development, per capita income is thought to
be negatively correlated with the risk of civil war onset (Fearon and Laitin 2003; Sambanis
2004; Collier, Hoe✏er and Rohner 2009; Bleany and Dimico 2011). That high-income
countries are less likely to experience civil war is the most “widely accepted relationship
between economic factors and civil war” (Dixon 2009, p.714). On the other hand, poorer
states with minimal resources at their disposal are thought to be more conflict-prone. In our
model, we control for the natural logarithm of a state’s gross domestic product per capita
using data from Heston, Summers and Aten (2012).
Second, we control for population size. Typically, scholars expect that states with a
higher population are more likely to experience civil conflict (Sambanis 2001; Collier and
Hoe✏er 2004; Fearon and Laitin 2003; Reynal-Querol 2002; Salehyan and Gleditsch 2006;
3
Bleany and Dimico (2011) analyze the pairwise correlation for war onsets across five
di↵erent datasets and find that it ranges from 0.197 to 0.634.
4
We use the 2010 updates from the COW dataset and the 2012 updates from the
UCDP/PRIO Armed Conflict Dataset.
15
Gleditsch 2007). In a subnational analysis of African states, Raleigh and Hegre (2007)
demonstrate that conflict incidence increases with population size and tends to erupt in
densely populated areas. We use the natural logarithm of a state’s population drawing on
data from Heston, Summers and Aten (2012).
A study of civil war onset must also control for regime type. Ellingsen and Gleditsch (1997) demonstrated a curvilinear relationship between democracy and conflict: highly
democratic and highly autocratic regimes are the most resistant to political instability. Following others (DeNardo (1985); Muller and Weede (1998); Regan and Henderson (2002)) we
expect anocratic states to be more likely to experience violent conflict. We therefore control
for the presence of anocracy using the square of a state’s Polity IV score, lagged by one year
(Gurr 1974; Marshall and Jaggers 2002).
Next, we control for the existence of ethnic grievances within a state. The extensive
literature on its connection to civil war onset remains divisive because so many metrics are
used to capture ethnic fractionalization, dominance, and violence. Early civil war models
included a measure of ethnolinguistic fractionalization (ELF) popularized by Easterly and
Levine (1997). Critics question its calculation and application in conflict models (Alesina
et al.
2003; Fearon 2003; Posner 2004), and several refined indices of ethnic division,
identity, and grievances have been developed (see Scarritt and Moza↵ar 1999; Roeder 2001;
Reynal-Querol 2002; Posner 2004; Chandra 2009; Wimmer, Cederman and Min 2009;
Chandra 2012) to better operationalize the mechanisms that fuel violence. Our model
employs an ethnic dominance measure from Salehyan and Gleditsch (2006), which captures
the percent of the population that does not belong to a dominant group, whether religious,
linguistic, or racial. Higher values, based on coding from Vanhanen (1999), indicate a smaller
16
dominant ethnic minority.5
Finally, because multiple civil war onset studies note time dependence issues, we follow
Carter and Signorino (2010) and use a cubic polynomial approximation. We opt for this
method over the time dummies or splines suggested by Beck, Katz and Tucker (1998) because
it avoids the quasi-complete separation problem.
Results: Civil Conflict Across Time and Space
Models 1 and 3 in Table 2 present baseline domestic models of civil war onset, using ACD
and COW data, respectively. Increasing population and ethnic heterogeneity have positive
and statistically significant e↵ects on the likelihood of conflict onset in both models. As GDP
per capita increases, the likelihood of conflict onset decreases when using UCDP’s definition
of conflict. Although this variable does not attain statistical significance in the COW model,
the direction of the e↵ect remains the same. The squared Polity Score captures the strength
of a regime (democracy or autocracy), with higher values indicating a more consolidated
regime. Model 3 confirms that a more consolidated regime (less anocratic) is less likely to
experience civil conflict onset. While the e↵ect is not significant in Model 1, it does attain
statistical significance in Model 3 (p<.01).
Next, Models 2 and 4 (ACD and COW, respectively) add the CE score to the baseline
models. Both specifications reveal that a more conflictual environment increases the likelihood of civil war onset. The e↵ects of the CE score are statistically significant (p<.001
for both the ACD and COW model). Models 2 and 4 indicate that including the CE score
improves the statistical model’s fit over that for models that exclusively rely on domestic
5
We run alternative models with Fearon and Laitin’s (2003) traditional ethnic fractional-
ization measure and Wimmer, Cederman, and Min’s (2009) excluded population measure.
The excluded population measure codes access to executive power, or the percentage of the
population excluded from executive positions.
17
Table 2: E↵ects of the Conflict Environment on Civil War Onset
(1)
ACD
(2)
ACD
(3)
COW
0.375⇤⇤⇤
(0.077)
Civil CE Score (ACD)
Civil CE Score (COW)
GDPpc (ln)
Population (ln)
Ethnic Heterogeneity
Squared Polity Score
Time
Time Squared
Time Cubed
Constant
AIC
BIC
Observations
(4)
COW
0.142⇤⇤
(0.045)
0.116⇤⇤
(0.037)
0.00664⇤⇤
(0.002)
0.00313
(0.002)
0.0400
(0.030)
0.00247
(0.002)
0.0000434
(0.000)
1.864⇤⇤⇤
(0.439)
930.9
984.1
5649
0.109⇤⇤
(0.042)
0.0845⇤
(0.033)
0.00545⇤
(0.002)
0.00363⇤
(0.002)
0.0453
(0.035)
0.00238
(0.002)
0.0000389
(0.000)
1.778⇤⇤⇤
(0.411)
906.5
966.1
5554
Standard errors in parentheses
Variables lagged one year (except neighborhood and time variables)
⇤
p < 0.05, ⇤⇤ p < 0.01, ⇤⇤⇤ p < 0.001
18
0.0835
(0.052)
0.1757⇤⇤⇤
(0.037)
0.00624⇤
(0.003)
0.00451⇤⇤
(0.002)
0.0234
(0.036)
0.00062
(0.002)
0.0000022
(0.000)
2.978⇤⇤⇤
(0.567)
1360.6
1413.7
5649
0.459⇤⇤⇤
(0.048)
0.0384
(0.054)
0.154⇤⇤⇤
(0.042)
0.00378
(0.003)
0.00435⇤
(0.002)
0.0308
(0.041)
0.00141
(0.002)
0.0000149
(0.000)
3.201⇤⇤⇤
(0.550)
1256.3
1315.9
5554
determinants of civil war onset.6
Using Model 2’s results and holding all domestic controls at their median values, Figure
2 plots the predicted probability of civil war onset as the CE score increases. As the CE
score (ACD specification) ranges from its minimum value to its maximum, the predicted
probability of civil war onset ranges from approximately one percent to close to twenty
percent. In very violent conflict environments, then, the likelihood of civil war onset is
substantially larger than those environments marked by peace.
Figure 2: Predicted Probabilities of Civil War Onset
Alternative Measures of Neighborhood E↵ects
While Table 2 reveals that the relationship between conflictual environments and civil war
onset is robust across civil war datasets, we acknowledge that researchers have used alternative measures of that environment. Therefore it seems prudent to ask whether our CE
6
A comparison of Model 2 to Model 1 (baseline) demonstrates that both the AIC and BIC
are lower for the conflict environment model; the same holds true when comparing Models
4 and 3.
19
score outperforms traditional measures of regional or neighborhood violence, especially since
other measures have produced mixed results, as we noted above.
First, we explore the e↵ects of neighborhood civil war when its operationalization depends
on contiguity. In their analysis of the determinants of civil war onset, Hegre and Sambanis
(2006) found the presence of civil war in an immediate neighbor to be one of the most robust
explanations for civil war onset. Consequently, this measure has been used in later studies
of civil war onset (see Buhaug and Gleditsch 2008; Braithwaite 2010a). We use Buhaug
and Gleditsch’s (2008) dichotomous variable to indicate the existence of a neighbor at war,
reconstructing their variable to be consistent with the most updated version of UCDP’s ACD
(Themner and Wallensteen 2012).
Second, we test a variable measuring the presence of regional conflict. If conflict di↵uses
via mechanisms beyond physical contagion, we expect that states with regionally-based cultural, political, and geographic ties would exhibit similar levels of conflict susceptibility.
Post-communist Eastern Europe during the 1990s and the Middle East/North African region during the late 2000s illustrate why we should not assume that conflict-ridden, noncontiguous states in the same region are independent. Therefore, we generate a variable to
capture the number of regional participants in civil conflict, basing our coding of region on
Fearon and Laitin’s work (2003) and the Minorities at Risk Project (Gurr et al. 2009).
Finally, we also consider how neighborhood economic conditions influence the onset of
civil war. Buhaug and Gleditsch (2008) suggest that neighborhood e↵ects on conflict occur because states with conflict-prone characteristics, such as poverty, are geographically
clustered. The conflict trap hypothesis popularized by Sambanis and co-authors (2003) suggests that relatively poor and politically unstable countries are spatially grouped. Following
Braithwaite (2010a), we include the average income of a state’s neighbors as a measure of
neighborhood income.
Table 3 replicates the analyses using each of the three alternative neighborhood measures
to the baseline model of civil war onset. Only the Regional States in Conflict (Model 6)
20
Table 3: Regional and Neighborhood E↵ects on Civil War Onset
(5)
ACD
Neighborhood Civil War
(6)
ACD
0.120
(0.103)
0.0407⇤
(0.020)
Regional States in Conflict
Neighborhood GDP (ln)
GDPpc (ln)
Population (ln)
Ethnic Heterogeneity
Squared Polity Score
Time
Time Squared
Time Cubed
Constant
AIC
BIC
Observations
(7)
ACD
0.150⇤⇤
(0.046)
0.122⇤⇤
(0.037)
0.00679⇤⇤
(0.002)
0.00318
(0.002)
0.0375
(0.030)
0.00245
(0.002)
0.0000436
(0.000)
1.805⇤⇤⇤
(0.439)
931.8
991.5
5649
0.0969
(0.055)
0.116⇤⇤
(0.036)
0.00670⇤⇤
(0.002)
0.00278
(0.002)
0.0434
(0.030)
0.00207
(0.002)
0.0000340
(0.000)
2.330⇤⇤⇤
(0.538)
928.4
988.1
5623
0.0769
(0.077)
0.111
(0.063)
0.112⇤⇤
(0.039)
0.00592⇤
(0.002)
0.00293
(0.002)
0.0225
(0.035)
0.00130
(0.002)
0.0000226
(0.000)
1.518⇤⇤
(0.524)
887.3
945.9
4938
Standard errors in parentheses
Variables lagged one year (except neighborhood and time variables)
⇤
p < 0.05, ⇤⇤ p < 0.01, ⇤⇤⇤ p < 0.001
coefficient is statistically significant (p<.05), but it is quite sensitive to model specification.
Lagging the variable and altering the set of control variables renders its e↵ects insignificant.
In Models 5 and 7, the coefficients for Neighborhood Civil War and Neighborhood GDP,
respectively, have the expected signs, but fail to attain statistical significance. The results
for traditional neighborhood measures found here mimic the mixed support uncovered in
the literature. We surmise that the indeterminate role of spatial di↵usion is an artifact of
21
indicators that do not take into account more complex spatial features of the surrounding
environment as well as temporal ones. As Model 2 from Table 2 demonstrated, our richer
and more dynamic measure of neighborhood context, the conflict environment, has a positive
and statistically significant e↵ect on conflict onset (p<0.001). Furthermore, a model using
the CE score generates better AIC and BIC values than the only one producing a statistically
significant coefficient for an alternative neighborhood measure (Model 6).
Additional Robustness Checks
To verify the robustness of our findings, we complete a number of checks which are available in the online appendix. Following Collier and Hoe✏er (2004), Hegre et al. (2001), and
Sambanis (2001, 2004), we explore an alternative coding of civil war onset that drops observations in which a civil war is ongoing. This removes 430 observations in the COW baseline
model and 549 observations in the ACD baseline model but our findings remain robust.
Second, we explore the e↵ect of alternative neighborhood variables on the Correlates of War
dataset. Third, we run alternative models on both civil war datasets that include di↵erent
operationalizations for ethnic diversity (Roeder 2001; Alesina et al.
2003; Wimmer, Ce-
derman and Min 2009), democracy (Marshall and Jaggers 2002), and political instability
(Fearon and Laitin 2003; Cederman, Wimmer and Min 2010). Fourth, we experiment with
di↵erent temporal lags for neighborhood variables to ensure that they do not become far
more statistically significant when lagged by only one or two years. Finally, we add our CE
score to models of civil war onset proposed by Buhaug and Gleditsch (2008) and Fearon and
Laitin (2003). Doing so improves the explanatory power of the model without substantially
changing the significance of other key variables. Ultimately, our robustness checks on the
spatial and temporal e↵ects of neighborhood conflict are consistent with our main empirical
analysis.
22
Conclusion
A state’s surroundings a↵ect its proclivity for civil war. That claim has made intuitive sense
to scholars for a long time, but supporting evidence has been mixed and fleeting until now.
Our improved conceptualization and measure of a state’s conflict environment fluctuates
and updates as new events, especially those that are more geographically proximate, in
that environment arise and as old ones fade away. Empirical analyses confirm that violent
conflict environments increase the likelihood of civil war onset; suggesting that both direct
and indirect di↵usion mechanisms spur new cases of civil war and that lingering memories
of violence contribute to its spread as well.
Additionally, our approach o↵ers researchers the flexibility to incorporate the conflict
environment history into an explanation of civil war that does not dismiss critical domestic
determinants of violence. This strikes a balance between recognizing and incorporating a
state’s surroundings and connections with the world and grounding civil war explanations
in domestic political processes.
Future research is warranted along both theoretical and empirical lines. For example,
more work is needed on the factors that condition the transmission of conflict across space.
In other words, how can scholars be smarter about when and how conflict in the neighborhood is expected to exacerbate the risk of violence at home? Environmental factors could
include transmission mechanisms such as road networks, border crossings, and refugee flows
(Krcmaric 2014). Future work must think theoretically about the internal factors, such
as economic and institutional characteristics, that make states vulnerable to environmental
pressures. In this analysis we assume that all states are equally susceptible to perturbations in their conflict environment, but we suspect this is an oversimplification. Mapping
conflict environments in conjunction with identifying the states that are most likely to be
a↵ected by their surroundings will improve our ability to predict the onset of civil wars.
Additionally, scholars should recognize that conflict is not the only important dimension of
a state’s environment. Other meaningful dimensions might be defined by neighbors’ regime
23
characteristics and communities or economic factors (Ahlquist and Wibbels 2012).
Finally, it is important to note that our temporal innovations in modeling the conflict
environment are intended to form a baseline model. Scholars who seek a more nuanced
historical context for a particular state or part of the world can build upon this to better
represent the way history influences decision-making with states. Some states have cultivated
an institutional sensitivity to nearby threats that may a↵ect the permanence of the impact
of historical violence. Serbians, for example, represent outside threats by memorializing the
fourteenth century Battle of Kosovo in poetry, song, and film. For Czechs it is the battle
of Bila Hora in 1620. Such chosen traumas can make states more susceptible to reacting to
neighboring violence with internal violence (Volkan 2011, 88-89). Quantifying the existence
and impact of these institutionalized memories would be difficult if not impossible on a
large-N scale, but these cases illustrates one of the many ways qualitative research can
improve upon baseline quantitative analyses and extend our conceptualization of conflict
environments.
24
References
Ahlquist, John S and Erik Wibbels. 2012. “Riding the Wave: World Trade and Factor-Based
Models of Democratization.” American Journal of Political Science 56(2):447–464.
Alcock, Norman Z. 1972. The War Disease. CPRI Press.
Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat and Romain
Wacziarg. 2003. “Fractionalization.” Journal of Economic Growth 8:155–194.
Ayres, William R. and Stephen Saideman. 2000. “Is Separatism as Contagious as the Common Cold or as Cancer? Testing International and Domestic Explanations.” Nationalism
and Ethnic Politics 63(3):91–113.
Beck, Nathaniel, Jonathan N. Katz and Richard Tucker. 1998. “Taking Time Seriously:
Time-Series Cross-Section Analysis with a Binary Dependent Variable.” American Journal
of Political Science 42(4):1260–1288.
Bleany, Michael and Arcangelo Dimico. 2011. “How di↵erent are the correlates of onset and
continuation of civil wars?” Journal of Peace Research 48(2):145–155.
Blimes, Randall J. 2006. “The Indirect E↵ect of Ethnic Heterogeneity on the Likelihood of
Civil War.” Journal of Conflict Resolution 50(4):536–547.
Bosker, Maarten and Joppe de Ree. 2014. “Ethnicity and the Spread of Civil War.” Journal
of Development Economics 108:206–221.
Braithwaite, Alex. 2010a. Conflict Hot Spots: Emergence, Causes, and Consequences. Ashgate.
Braithwaite, Alex. 2010b. “Resisting Infection: How State Capacity Conditions Conflict
Contagion.” Journal of Peace Research 47(3):311–319.
25
Bremer, Stuart. 1982. “The Contagiousness of Coercion: The Spread of Serious International
Disputes, 1900-1976.” International Interactions 9(1):29–55.
Buhaug, Halvard and Kristian Skrede Gleditsch. 2008. “Contagion or Confusion? Why
Conflicts Cluster in Space.” International Studies Quarterly 52:215–233.
Call, Charles T. 2011. “Beyond the ’Failed State’: Toward Conceptual Alternatives.” European Journal of International Relations 17(2):303–326.
Carter, David B. and Curtis S. Signorino. 2010. “Back to the Future: Modeling Time
Dependence in Binary Data.” Political Analysis 18(3):271–292.
Cederman, Lars-Erik, Andreas Wimmer and Brian Min. 2010. “Why Do Ethnic Groups
Rebel? New data and analysis.” World Politics 62(1):87–119.
Cederman, Lars-Erik, Kristian Skrede Gleditsch and Halvard Buhaug. 2013. Inequality,
Grievances, and Civil War. Cambridge Univeristy Press.
Cederman, Lars-Erik, Kristian Skrede Gleditsch, Idean Salehyan and Julian Wucherpfennig.
2013. “Transborder Ethnic Kin and Civil War.” International Organization 67:389–410.
Cederman, Lars-Erik, Luc Girardin and Kristian Skrede Gleditsch. 2009. “Ethnonationalist
Triads: Assessing the Influence of Kin Groups on Civil Wars.” World Politics 61(3):403–
437.
Chandra, Kanchan. 2009. A Constructivist Dataset on Ethnicity and Institutions (CDEI).
In Identity as a Variable, ed. Rawi Abelal, Yoshiko Herrera, Ian Johnston and Rose McDermott. Cambridge University Press chapter 8.
Chandra, Kanchan, ed. 2012. Constructivist Theories of Ethnic Politics. Oxford University
Press.
Collier, Paul. 2000. “Rebellion as a Quasi-Criminal Activity.” Journal of Conflict Resolution
44:839–853.
26
Collier, Paul and A. Hoe✏er. 2004. “Greed and Grievance in Civil War.” Oxford Economic
Papers 56:563–595.
Collier, Paul, Anke Hoe✏er and Dominic Rohner. 2009. “Beyond Greed and Grievance:
Feasibility and Civil War.” Oxford Economic Papers 61:1–27.
Cranmer, Skyler J and Bruce A Desmarais. 2011. “Inferential Network Analysis with Exponential Random Graph Models.” Political Analysis 19(1):66–86.
Crescenzi, Mark and Andrew Enterline. 2001. “Time Remembered: A Dynamic Model of
Interstate Interaction.” International Studies Quarterly 45(3).
Danneman, Nathan and Emily Hencken Ritter. 2014. “Contagious Rebellion and Preemptive
Repression.” Journal of Conflict Resolution 58(2):254–279.
DeNardo, James. 1985. Power in Numbers: The Political Strategy of Protest and Rebellion.
Princeton University Press.
DiCarlo, Rosemary A., ed. January 26, 2012. Remarks by Ambassador Rosemary DiCarlo,
U.S. Deputy Permanent Representative to the United Nations, at Security Council Briefing
on Peace and Security in Africa. United Nations.
Dixon, Je↵rey. 2009. “What Causes Civil Wars? Integrating Quantitative Research Findings.” International Studies Review 11(4):707–735.
Dor↵, Cassy and Michael D Ward. 2013. “Networks, Dyads, and the Social Relations Model.”
Political Science Research and Methods 1(02):159–178.
Easterly, William and Ross Levine. 1997. “Africa’s Growth Tragedy: Policies and Ethnic
Divisions.” Quarterly Journal of Economics 112:1203–1250.
Elbadawi, Ibrahim and Nicholas Sambanis. 2002. “How Much War Will we see? Explaining
the Prevalence of Civil War.” Journal of Conflict Resolution 46(3):307–334.
27
Ellingsen, Tanja and Nils Petter Gleditsch. 1997. Democracy and Armed Conflict in the
Third World. In Causes of Conflict in Third World Countries, ed. Ketil Volden and Dan
Smith. International Peace Research Institute pp. 69–81.
Fearon, James. 2003. “Ethnic Structure and Cultural Diversity by Country.” Journal of
Economic Growth 8:195–222.
Fearon, James D. and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American Political Science Review 97(1):75–90.
Fearon, James D., Kimuli Kasara and David D. Laitin. 2007. “Ethnic Minority Rule and
Civil War Onset.” American Political Science Review 101(1):187–193.
Forsberg, Erika. 2014a. “Di↵usion in the Study of Civil Wars: A Cautionary Tale.” International Studies Review 16:188–198.
Forsberg, Erika. 2014b. “Transnational Transmitters: Ethnic Kinship Ties and Conflict
Contagion 1946-2009.” International Interactions 40(2):143–165.
Gates, Scott, Havard Hegre, Mark P. Jones and Halvard Strand. 2006. “Institutional Inconsistency and Political Instability: Polity Duration, 1800-2000.” American Journal of
Political Science 50(4):893–908.
Gleditsch, Kristian Skrede. 2002. All International Politics is Local: The Di↵usion of Conflict, Integration, and Democratization. University of Michigan Press.
Gleditsch, Kristian Skrede. 2007. “Transnational Dimensions of Civil War.” Journal of Peace
Research 44(3):293–309.
Gleditsch, Kristian Skrede and Michael D. Ward. 2001. “Measuring Space: A MinimumDistance Database and Applications to International Studies.” Journal of Peace Research
38(6):739–758.
28
Gleditsch, Nils, Havard Strand, Mikael Eriksson, Margareta Sollenberg and Peter Wallensteen. 2002. “Armed Conflict 1946-2001: A New Dataset.” Journal of Peace Research
39(5):615–637.
Gurr et al. 2009.
“Minorities at Risk Dataset.” College Park,
MD: Cen-
ter for International Development and Conflict Management. Retrieved from
http://www.cidcm.umd.edu/mar/.
Gurr, Ted Robert. 1970. Why Men Rebel. Princeton University Press.
Gurr, Ted Robert. 1974. “Persistence and Change in Political Systems, 1800-1971.” American
Journal of Political Science 74:1482–1504.
Hegre, Havard and Nicholas Sambanis. 2006. “Sensitivity Analysis of Empirical Results on
Civil War Onset.” Journal of Conflict Resolution 50(4):508–535.
Hegre, Havard, Tanja Ellingsen, Scott Gates and Nils Petter Gleditsch. 2001. “Towards a
democratic civil peace?” American Political Science Review 95:33–48.
Heston, Alan, Robert Summers and Bettina Aten. 2012. “Penn World Table Version 7.1,
Center for International Comparisons of Production, Income and Prices at the University
of Pennsylvania.”.
Hill, Stuart, Donald Rothchild and Colin Cameron. 1998. Tactical Information and the Di↵usion of Peaceful Protests. In The International Spread of Ethnic Conflict: Fear, Di↵usion,
and Escalation, ed. David A. Lake and Donald Rothchild. Princeton University Press.
Houweling, Henk and Jan Siccama. 1985. “The Epidemiology of War, 1816-1980.” Journal
of Conflict Resolution 29(4):641–663.
Kadera, Kelly. 1998. “Transmissions, Barriers, and Constraints: A Dynamic Model of the
Spread of War.” Journal of Conflict Resolution 42(3):367–387.
29
Ki-Moon, Ban. 2012. Report of the assessment mission on the imapct of the Libyan crisis
on the Sahel region. Technical report United Nations Security Council.
Krcmaric, Daniel. 2014. “Refugee Flows, Ethnic Power Relations, and the Spread of Conflict.” Security Studies 23:182–216.
Kuran, Timor. 1998. Ethnic Dissimilation and its International Di↵usion. In The International Spread of Ethnic Conflict: Fear, Di↵usion, and Escalation, ed. David A. Lake and
Donald Rothchild. Princeton University Press.
Lacina, Bethany. 2004. “From Side Show to Centre Stage: Civil Conflict after the Cold
War.” Security Dialogue 35:191–2005.
Marshall, Monty G. and Keith Jaggers. 2002.
“Polity IV Project: Political Regime
Characteristics and Transitions, 1800-2002.” College Park, MD: Center for International Development and Conflict Management, University of Maryland. URL:
http://www.cidcm.umd.edu/inscr/polity/index.htm.
Most, Benjamin A. and Harvey Starr. 1980. “Di↵usion, Reinforcement, Geopolitics, and the
Spread of War.” American Political Science Review 74(4):932–946.
Muller, Edward N. and Erich Weede. 1998. “Rebellion, Violence and Revolution: A Rational
Choice Perspective.” Journal of Peace Research 35:43–59.
Murdoch, James C. and Todd Sandler. 2002. “Economic Growth, Civil Wars, and Spatial
Spillovers.” Journal of Conflict Resolution 46(1):91–110.
Posner, Daniel N. 2004. “Measuring Ethnic Fractionalization in Africa.” American Journal
of Political Science 48(4):849–863.
Raleigh, Clionadh and Havard Hegre. 2007. “Population Size, Concentration, And Civil
War : A Geographically Disaggregated Analysis.” Policy Research Working Paper from
the World Bank.
30
Regan, Patrick and Daniel Norton. 2005. “Greed, Grievance, and Mobilization in Civil
Wars.” Journal of Conflict Resolution 49(3):319–336.
Regan, Patrick M. and Errol A. Henderson. 2002. “Democracy, Threats, and Political Repression in Developing Countries: Are Democracies Internally Less Violent?” Third World
Quarterly 23(1):119–136.
Reynal-Querol, Marta. 2002. “Ethnicity, Political Systems, and Civil Wars.” Journal of
Conflict Resolution 46(1):29–54.
Roeder, Philip G. 2001. “Ethnolinguistic Fractionalization (ELF) Indices, 1961.”.
Rotberg, Robert. 2004. When States Fail: Causes and Consequences. Princeton University
Press.
Saideman, Stephen and R. William Ayres. 2000. “Determining the Causes of Irredentism:
Logit Analyses of Minorities at Risk Data from the 1980s and 1990s.” Journal of Politics
62(4):1126–1144.
Salehyan, Idean. 2008. “The Externalities of Civil Strife: Refugees as a source of international
conflict.” American Journal of Political Science 52(4):787–801.
Salehyan, Idean and Kristian Skrede Gleditsch. 2006. “Refugees and the Spread of Civil
War.” International Organization 60:335–366.
Sambanis, Nicholas. 2001. “Do Ethnic and Nonethnic Civil Wars Have the Same Causes? A
Theoretical and Empirical Inquiry.” Journal of Conflict Resolution 45:259–282.
Sambanis, Nicholas. 2004. “What is a Civil War? Conceptual and Empirical Complexities.”
Journal of Conflict Resolution 48:814–858.
Sambanis, Nicholas, Marta Reynal-Querol, Havard Hegre, Lani Elliott, Paul Collier and
Anke Hoe✏er. 2003. Breaking the Conflict Trap: Civil War and Development Policy. :
New Policy Research Report. World Bank and Oxford University Press.
31
Sarkees, Meredith Reid and Frank Wayman. 2010. Resort to War: 1816-2007. CQ Press.
Scarritt, James R. and Shaheen Moza↵ar. 1999. “The Specification of Ethnic Cleavages and
Ethnopolitical Groups for the Analysis of Democratic Competition in Africa.” Nationalism
and Ethnic Politics 5:82–117.
Singer, David J. and Melvin Small. 1982. Resort to Arms: International and Civil Wars.
Sage.
Siverson, Randolph M. and Harvey Starr. 1990. “Opportunity, Willingness, and the Di↵usion
of War.” American Political Science Review 84(1):47–67.
Stewart, Frances. 2002. Horizontal Inequalities as a Source of Conflict. In From Reaction to
Conflict Prevention: Opportunities for the UN System, ed. Fen Olser Hampson and David
Malone. Lynne Rienner.
Themner, Lotta and Peter Wallensteen. 2012. “Armed Conflict, 1946-2011.” Journal of
Peace Research 49(4).
Tufekci, Zeynep and Christopher Wilson. 2012. “Social Media and the Decision to Participate in Political Protest: Observations From Tahrir Square.” Journal of Communication
62(2):363–379.
Vanhanen, Tatu. 1999. “Domestic Ethnic Conflict and Ethnic Nepotism: A Comparative
Analysis.” Journal of Peace Research 36(1):55–74.
Volkan, Vamik D. 2011. “Transgenerational Transmissions and Chosen Traumas: An Aspect
of Large-Group Identity.” Group Analysis 34(1):79–97.
Ward, Michael D. and Kristian Skrede Gleditsch. 2008. Spatial Regression Models: Quantitative Applications in the Social Sciences. Sage Publications.
Weidmann, Nils B., Doreen Kuse and Kristian Skrede Gleditsch. 2010. “The Geography of
the International System: The CShapes Datset.” International Interactions 36(1).
32
Wimmer, Andreas, Lars-Erik Cederman and Brian Min. 2009. “Ethnic Politics and Armed
Conflict: A Configurational Analysis of a New Global Dataset.” American Sociological
Review 74(2):316–337.
33